| title | Pixelated Empathy Documentation |
|---|---|
| description | Documentation for Pixelated Empathy — the clinical mental health AI platform that trains professionals through simulated therapeutic practice. |
Pixelated Empathy is a clinical mental health AI platform. It provides training infrastructure for mental health professionals to practise difficult therapeutic conversations against AI clients, with real-time feedback on empathy, technique, bias, and crisis response.
The platform is not a chatbot or a therapy replacement. It is empathy infrastructure — a gym where clinicians build and refine therapeutic skills through repeated, scored practice.
Pixelated Empathy is a proprietary, Early Access product. Visit [pixelatedempathy.com](https://pixelatedempathy.com) for access.Step-by-step tutorials for therapists, supervisors, and developers. Learn by doing. How-to guides for working with the platform in real scenarios. API documentation, architecture references, compliance, and enterprise materials. The psychological frameworks, scoring models, and AI systems behind the platform. Documentation for the three components: Pixelated, the AI engine, and Foresight.
Pixelated Empathy consists of three interconnected components:
| Component | Role | Stack |
|---|---|---|
| Pixelated | Main orchestrator — app surfaces, clinical dashboards, API routes | Astro, React 19, TypeScript, Express, Python |
| AI Engine | Cognitive engine — model training, inference, emotional signal analysis | Python, PyTorch, SFT/DPO/GRPO |
| Foresight | Continuity and memory — persistent memory for AI agents | Python, PostgreSQL, MCP server |
Each component has its own repository and product overview under Products.
- FHE-Based Security — Strict HIPAA compliance via Fully Homomorphic Encryption. Patient data remains encrypted even during model inference.
- The Empathy Gym — 410+ validated therapeutic scenarios across 15+ cultural contexts with real-time feedback.
- Real-Time Bias Detection — 40+ bias categories monitored during sessions with equitable AI guardrails.
- Crisis Detection — Five signal categories with five-level risk classification and escalation protocols.
- Foresight Continuity — Longitudinal memory via MCP architecture for multi-session therapeutic arcs.
Practise therapeutic conversations against AI clients with scored feedback on empathy, technique, and safety. Configure training programmes, monitor trainee progress, and set competency thresholds. Set up the platform locally, integrate via API, or extend the cognitive engine. Enterprise deployment, compliance documentation, SLAs, and vendor management.